A Minimalist Approach to the Null Subject Parameter

Abstract Null Subject Romance Languages are languages in which the V feature of T is necessarily strong. Taking so much for granted, and given the morphological characteristics of this kind of languages, all their properties follow: V raising deposits at T a set of [– interpretable] φ-features that are erased through covert raising by the [+interpretable] φ−features of the subject DP, either phonetically realized or pro (pro: a notational abbreviation for a set of features composed by the categorial feature D and φ-features). Inverted subject (DP) and null subject (pro) are, therefore, vP internal subjects. The preverbal subject, in contrast, is a DP merged at T —therefore a non-argument— in a derivation with a pro as the internal subject. Starting from this basic proposal, the analysis of some new data points to the spurious character of Nominative Case in these languages and suggests that pro is only allowed in Spec, vP. Key words: Null Subject Languages, [–interpretable] V-features, derivational approach, pro, Nominative. Resum. Un enfocament minimalista del paràmetre del subjecte nul Les llengües de subjecte nul romàniques, com ara el català, l’espanyol i l’italià, són llengües que tenen el tret V de T necessàriament fort. Donant això per descomptat i ateses les característiques morfològiques de la concordança verbal d’aquest tipus de llengües, el conjunt de propietats que les caracteritzen se’n segueix automàticament: l’elevació de V diposita a T un joc de trets φ [–interpretable] que són esborrats, mitjançant trasllat no manifest, pels trets φ[+interpretable] de l’SD subjecte, fonèticament realitzat o pro (pro: una abreviatura notacional per a un a conjunt de trets compost del tret categorial D i trets φ). Subjecte invertit (SD) i subjecte nul (pro) són, doncs, subjectes interns a l’Sv. El subjecte preverbal, en canvi, és un SD ajuntat a T —I, doncs, un no argument— en una derivació amb pro com subjecte intern. Aplicant aquesta proposta bàsica a l’anàlisi d’algunes dades empíriques noves, s’apunta el caràcter espuri del cas nominatiu en aquestes llengües i se suggereix que pro només és permès a l’Espec d’Sv. Paraules clau: Llengües de Subjecte Nul, trets de V [–interpretable], enfocament derivacional, pro, nominatiu. * I am grateful to the audience of Seminari de Gramàtica Teòrica of the Universitat Autònoma de Barcelona for their comments on a previous version of this paper. Thanks are due specially to CatWPL 8, 2000 97-128 A Minimalist Approach to the Null Subject Parameter* Joana Rosselló Universitat de Barcelona. Departament de Lingüística General. Gran Via de les Corts Catalanes, 585. 08007 Barcelona. [email protected] Received: November 20th 2000 Accepted: Desember 14th 2001 CatWPL 8 001-137 29/5/2001 17:30 Página 97 1. Introduction As far as I know, there is yet no truly minimalist proposal accounting for the basic properties displayed by Null Subject Languages (NSLs),1namely those stated in (1). Such a proposal should be able to combine two achievements: on the theoryinternal side, it should not appeal to more than one Infl category and it should be strictly derivational; on the empirical side, it should account at least for the three properties in (1) without resort to any mechanism which is extraneous to the (human) language Computational System (CSHL). (1) a. Preverbal Subject En Joan ballava. the Joan dance-PAST-3SG ‘Joan danced.’ b. Postverbal Subject Ballava en Joan. dance-PAST-3SG the Joan ‘Joan danced.’ c. Null Subject Ballava. dance-PAST-3SG/1SG ‘He/she/I danced.’ Let us begin by considering how the current accounts fail to achieve these two goals. On the one hand, it has to be pointed out that there is no proposal being able to explain the properties in (1), this failure being due either to omission —as in Barbosa (1997), Manzini & Savoia (1997), Ordóñez & Treviño (1999) and Pollock 98 CatWPL 8, 2000 Joana Rosselló Jaume Solà for his invaluable help in the preparation of the English version. The research that underlies this work has been supported by grant PB96-1199-C04-02 from the Ministerio de Educación y Cultura and by grant 1999SGR00113 from the Generalitat de Catalunya. 1. I want to point out that in this paper the expressions «subject», «preverbal subject», «Null Subject Parameter» etc. are used in a rather informal way, and that by «Null Subject» languages (NSLs) we intend those in which the non-realisation of the subject as a phonetically overt DP is possible because of verbal agreement. We therefore exclude Chinese-style NSLs, and include both NSLs with uniform agreement (NSLsUA) —such as Catalan, Spanish or Italian— and NSLs with alternate agreement (NSLsAA) —such as Arabic, Celtic languages, etc. Table of Contents 1. Introduction 2. The verbal φ-features in NSLs 3. Towards a minimalist approach 4. The EPP feature of T is not universal 5. A minimalist approach to the Null Subject Parameter 6. Impossible preverbal (or null) subjects 7. Conclusions References CatWPL 8 001-137 29/5/2001 17:30 Página 98 (1997)—, or to the conclusion that the properties in our triad (1a)-(1c) are not derivable from the mechanisms of CSHL —as in Picallo (1998) and Donati & Tomaselli (1997). So, in Picallo (1998), since the subjet preposing operation that would yield (1a) constitutes an economy violation when compared with (1b), it could not be the result of operations driven by CSHL to satisfy feature checking, but the result of a discourse-driven operation.2 On the other hand, we are dealing with approaches which are neither completely nor truly derivational: there is often a mixture of representational and derivational criteria. So, for instance, when it is said that the preverbal subject is dislocated (in the sense of sentence-external), it remains a mystery how that kind of subject has been derived and, thus, whether it enters feature checking, if it has to. More generally, the lack of explicitness about the derivations underlying each construction extends to all the members of the triad. Finally, it seems worth mentioning that in almost all the references on this topic appeal is made to more than one inflectional category. So, some of them posit AGRP and TP (Alexiadou & Anagnostoupoulou (1996, 1998), Benedicto (1993), Nash & Rouveret (1996), Platzack (1994), Donati & Tomaselli (1997)); and some others posit two IPs with different features (Manzini & Savoia (1997), Cardinaletti (1994, 1997)). Taking seriously the state of affairs this survey shows, it appears as more than just expedient to pave the way for an account according to the minimalist guidelines cited above. In particular, I will develop a system in which the [–interpretable] features play a key role in the derivation. In what follows, I will show that the trigger for the properties listed in (1a)- (1c) is the bunch of [–interpretable] φ−features of the tensed verb in NSLs (section 2). Then (section 3), I will present the basic guidelines of my proposal. One of its results (section 4) concerns the non-universality of the EPP, which I pursue, while arguing, however, against the current proposals according to which there is no Spec,T position in NSLs with Uniform Agreement (NSLUA) —see note 1. Against these views, a new set of data is presented which obliges us to be indulgent with the EPP in its classical version, this is to say, as obligatory presence of Spec,T. A meaningful and, until now, unnoticed contrast between English and A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 99 2. Picallo (1998) assumes that the same numeration underlies (1a) and (1b). As we will see (§ 4.1), it can be argued that this assumption is mistaken. Donati & Tomaselli (1997) conclude that neither the Extended Projection Principle (EPP), nor the Null Subject Parameter can be derived under the elementary mechanisms of CSHL. The fact that their work crucially relies on two inflectional heads —a handicap to be avoided in the present proposal— allows us to leave it aside without having to deal with the arguments that lead them to such a conclusion. Nevertheless it must be noticed that Donati and Tomaselli aim at explaining the syntactic behaviour of subjects in English, French and German; pro-drop languages —our NSLUA—, and VS languages such as Irish —an instance of NSLAA. Unlike all the proposals mentioned so far, Kato’s (1999) does not incur any of these problems. But it assumes that verbal agreement is categorially a D(eterminer) and, as such, an independent lexical item which enters the derivation as a true argument. The problems with such an approach will be addressed later in the text (§ 2, 4.1). CatWPL 8 001-137 29/5/2001 17:30 Página 99 Catalan is offered that shows it is mistaken to assume that, beyond the fact that English has no pro, in both languages there are preverbal subjects (arguments or expletives) of the same kind. In the following section (section 5), the basic proposal already sketched in section 3 is presented in a more explicit way and compared with the classical ones, namely those assuming that Spec,T is always filled. Merge with T (or above T) yielding the preverbal subject in NSLs is also examined. Finally (section 6), two almost unnoticed contrasts between preverbal and postverbal subjects lead me to explore two issues: the possibility that Nominative Case is spurious in NSLs, and the constrained distribution of pro, that would only appear in Spec,vP. Such an approach to Nominative Case yields a new insight into the concurrence of Partitive Case and agreement (with the verb) and makes it possible to establish a clear separation between clitics and agreement affixes. To conclude this introduction, a warning note is in order. Given the limitations of this paper, it would be impossible to deal with two more issues that constitute, in fact, the background of my proposal. The first one, that I take for granted, is the left-dislocated status of preverbal subjects in NSLs (see Solà 1992, Ordóñez & Treviño 1999, Alexiadou & Anagnostoupoulou 1998), although in the account to be presented here it simply means that the preverbal subject is a non argument that, in a similar way to expletives, merges in a non-θ-position, in fact, often in the same position the θ-subject raises to in non-NSLs (see section 5). The second issue is that the purported evidence that pro is preverbal (in the sense that it appears in Spec,T) —see, for instance, Rizzi (1990), Cardinaletti (1994, 1997)— is much less compelling than it is usually assumed to be: parallelism between sentences with pro as subject and sentences with a lexical preverbal subject is also expected in a framework where lexical (and non focused) preverbal subjects are nothing but sentences with a null subject (pro) in θ-position and an (optionally added) left-dislocated subject, in fact a non-argument merged in a non-θ-position. With these two remarks in mind, we henceforth concentrate on the proposal to account for the properties in (1a)-(1c).3 2. The verbal φ-features in NSLs The present proposal, as all those in the past concerning NSLs, relates the properties in (1) to the «richness»4of verbal agreement. So far, therefore, nothing is new. Since Taraldsen (1980) and Rizzi (1982), all the proposals resort, in similar ways, to «rich agreement». In the REST framework (Chomsky, 1981), this guiding idea was fleshed out as INFL having the feature [+pronominal]. As far as I know, how100 CatWPL 8, 2000 Joana Rosselló 3. I leave out of the present paper showing how the behaviour of copulative sentences in NSLsUA, can be considerably clarified within the framework advocated here. As is well known, this kind of construction shows agreement to the right: (i) Sóc jo. am I ‘It’s me.’ 4. On how richness is to be understood, see Kato (1999), who systematises previous proposals. CatWPL 8 001-137 29/5/2001 17:30 Página 100 ever, there is as yet no proposal in which the [–interpretable] verbal features (number, (gender), (person)), when so characterized, play the role they are assumed to play in the workings of NSLs, it is to say, in triggering displacement. Thus, Kato (1999) and Ordóñez & Treviño (1999), for instance, give argument status to the affix containing the [–interpretable] features of V, instead of considering it a reply of the [+interpretable] φ-features of the corresponding DP argument. In this way, the φ-features of V become, ipso facto, [+interpretable]. Alexiadou & Anagnostoupoulou (1998), and Pollock (1997) do not see the affix as an argument, but they also assume (although not explicitly) that the φ-features of V are [+interpretable]: they must indeed be when the affix is taken to belong to the category of pronouns, as they propose (see Alexiadou & Anagnostoupoulou 1998: 517). Since they consider the agreement affix a pronoun with φ-features, it turns out that V raising to T checks not only the V-feature of T but also the D-feature of T (EPP), i.e., the latter checking takes place by adjunction to the head (T) instead of substitution in Spec,T. Independently of the empirical problems tied with such views —see §6—, it must be pointed out that on the theory-internal side it is indeed impossible to think of φ-features of V as [+interpretable]. The reason is that to achieve this interpretation one must reject the Strong Lexicalist Hypothesis (SLH), which seems to be inherent to Minimalist Program (MP). Indeed, according to SLH, the array of lexical choices provided by the numeration are lexical items already inflected that provide each choice with an unique category.5This being so, there is no way to escape the conclusion that φ-features of V are [–interpretable]: to achieve the opposite conclusion it would be necessary to assume that these features are in a argument and this, in turn, would lead to introducing V and the agreement affix separately, which is just what the SLH prohibits. There are, moreover, general grounds for casting doubts on the accounts that grant inflectional affixes (clitics are to be kept apart) the status of an argument. Were they arguments, it should be explained why Alternate Agreement (AA) languages (see note 1), as Arabic or Celtic ones, work as defined by Greenberg’s Universal 33 and not conversely. Let us see why. Universal 33 states that «When number agreement between the noun and verb is suspended and the rule is based on order, the case is always one in which the verb is in the singular.» This statement, translated to our terms, amounts to saying that when the agreement between verb and subject is suspended, it is the verb that loses (all the) φ-features, whereas the φ-features of the subject DP remain unchanged. Note that if the inflectional affix was an argument, one could expect languages suspending the agreement between verb and subject the other way round, that is, having the verb with all the φ-features and the DP without. As an illustration, we should have languages with the pattern A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 101 5. Notice that for Pollock (1997), Alexiadou & Anagnostoupoulou (1998), etc., the verbal form would not only be V, but also D. This interpretation is, in my view, incompatible with the SLH, since it would make available to computation the lexical, categorial, inner parts of what in syntax counts as a single category —conceding it is legitimate, in the first place, to postulate the D category on the basis of mere φ-features. CatWPL 8 001-137 29/5/2001 17:30 Página 101 Work-3PL the man meaning ‘The men work’, instead of the existing pattern in this kind of AA languages, namely Work-∅the men. As far as I know, this behaviour is not attested across languages. We conclude, therefore, that the φ-features which enter into the subject verb agreement are primarily features of the DP subject (in which they are [+interpretable]) and, therefore, that it is the verb indeed that agrees with the subject, and not the subject whit the verb, as ever stated in good traditional grammar.6 3. Towards a minimalist approach Given the previous theory-internal considerations and this last empirical observation, it seems that one can take as well founded the main assumption of the present proposal, namely the [–interpretable] nature of the φ-features of V. Now, if V remains in the position in which it is merged initially, these features will not be able to attract the φ-features of SUBJ in the Spec,vP basic position, simply because here SUBJ stands above V. It is then necessary that V raises overtly, as in fact it does in all NSLs.7We then have a strong V feature of T and a set of [–interpretable] φ-features on the verb, raising along with it to T. What erases them?8In the present proposal —where the preverbal subject, as we will show, is not a true argument—, erasure of such features has to be accomplished covertly by the φ-features of the so called postverbal subject (the true internal subject in the case of unergative or transitive verbs, or the DP sister to V in unaccusative verbs). Hence, such an 102 CatWPL 8, 2000 Joana Rosselló 6. There is a certain idealisation in this argumentation, if we consider cases like: (i) Ho hem preparat els ajudants. it-have-1PL prepared the assistants ‘We the assistant professors have prepared it.’ (ii) Els nens encara no ho sabeu. the children still not it-know-2PL ‘You children don’t know yet.’ or the examples in (18). In fact, in these examples there is more information in the verbal ending (extra person information) than in the triggering DP. Whatever the precise analysis for these cases, though, we can reasonably maintain the argument in the text, however qualified, on the grounds that a mismatch in number (V-plural / DP-singular) is never attested. 7. It can be taken as an empirical fact that in NSLs V raising is overt. This is shown by, among other facts, adverb distribution (La Maria fa sovint la migdiada, the Maria does often the siesta). Is that, though, what we should expect, considering the [±interpretable] status of the features entering the computation? It seems the answer is yes. Indeed, if we assume, as seems reasonable, that the V-feature of T is [+interpretable], a verb which has not raised overtly would be left with no [–interpretable] attractor to get raised covertly, so that its [–interpretable] φ-features could not possibly be erased and the derivation would crash at the C-I interface. It would follow, then, that the presence of verbal agreement in NSLs must be associated to a strong V-feature of T. Following this line of reasoning, we could even conclude that attributing strength to a/some features(s) is not anymore a mere stipulation, but rather a consequence of the interplay between the [±interpretable] features within the system. See footnote 19 for further information. 8. Chomsky (1998) attributes φ-features to T itself. See, however, the following discussion in the text and footnote 19. CatWPL 8 001-137 29/5/2001 17:30 Página 102 erasure is driven by bare attraction of the φ-features of the DP, that adjoin to the [T V+φ-T] head, without pied-piping of the DP. I will now try to lay the foundations of the present account by showing the shortcomings of an alternative way of erasing the φ-features of V. This possibility, explored in Chomsky (1998), assumes that T is supplied itself with φ-features along with V and D/N-features. Then, depending on whether the V feature of T is (a) strong or (b) weak, we will have the following. If (a), the φ-features of V could be erased as free riders when V is attracted by the strong V-feature of T. Naturally, also the φ-features of T would be erased in this way. If (b), and if the D-feature of T is strong (EPP), the [–interpretable] φ-features of T would be erased by the corresponding [+interpretable] φ-features of DP, raised as free riders along with the DP attracted by the strong D-feature of T. The [–interpretable] φ-features of V would remain yet to be erased, and would have to raise as free riders along with the V feature, in an instance of pure feature-raising, without pied-piping of the whole category. Once in T-adjoined position, they would finally be erased by the φ-features of the DP in Spec,T. So far, (b) — which could apply to languages like English— would yield the intended result, namely the erasure of the φ-features of V by means of the corresponding ones in the DP in Spec,T. Still, there would be an apparently superfluous operation: what would the checking (and erasure) of the φ-features of T be used for? In contrast, in case (a) —which would apply to any language with overt V raising—, the result turns out to be unwelcome, independently of the strength of the D-feature of T, since, even in the case that D is strong, the Extension Condition requires that the raising of V precedes the overt raising of the DP subject, and so the checking (and erasure) of φ-features of V (and T, in fact) takes place in the configuration of adjunction to the head T. Indeed, this result is unsatisfactory not only with respect to the intended goal (only the features of a DP should be able to check the associated φ-features of V) but also with respect to the empirical evidence: by allowing erasure of φ-features of V by the φ-features of T, we should expect, contrary to fact, any feature mismatch between verb and subject to be possible.9 Given this empirical shortcoming, it seems that the assumption that T has its own φ-features must be turned down, and that we can keep to the idea that T has only Vand D/N-features. So φ-features would always be a property of lexical items. Upon reflection, positing φ-features for T is reminiscent of positing an AGR category, which, as is well known, was finally eliminated because it was not interpretable at the C-I interface and moreover consisted solely of φ-features and strength (see Chomsky, 1995, § 4.10). Furthermore, it should be noticed that, by eliminating the possibility of φ-features in T, case (b) behaves as desired: the φfeatures of DP will check (and erase) the associated φ-features of V without resort to the superfluous operation mentioned above. A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 103 9. See note 6. CatWPL 8 001-137 29/5/2001 17:30 Página 103 One furthermore can speculate on the following lines: conceptually, it is not very elegant that [–interpretable] features are checked (and erased) by also [–interpretable] features. Keeping to the minimalist guideline according to which [–interpretable] features are responsible of displacement, it seems more desirable that [–interpretable] features attract [+ interpretable] associated features.10 In any case, the present proposal adheres this kind of view. To summarize, the φ-features of V can only be erased by the features of a DP. In particular, in NSLsUA these features are erased by the corresponding DP argument (either overt o pro) within vP. So erasure is always accomplished by bare features, as an instance of head adjunction without pied-piping. In fact, this is straightaway suggested by (1b), i.e. by the possibility of a postverbal subject, and is required to derive the null subject (pro) sentences like (1c). In effect, on minimalist assumptions to raise pro overtly in an instance of substitution in Spec,T is not at all a sound move, since pied-piping applies only when it is required by the A-P interface. Keeping to minimalist guidelines, pro has to be considered a set of bare features, D and φ-features. So far, the possibility of an expletive pro (proEXPL) has not been yet ruled out. Such an element, if existing, would have to merge in the inflectional phase, given its non-argument nature. That proEXPL does not exist will be argued for in the next section. Suppose now that this argumentation is correct. In this case, a sentence with a postverbal subject would yield an EPP violation (EPP understood as checking/erasure of D-feature of T by means of a DP in Spec,T); furthermore, at first glance it would seem that in this way we preclude (as Picallo 1998 does) the possibility that the bare workings of CSHL can account for the double possibility concerning the subject position, either preverbal or postverbal. Without proEXPL the numeration underlying sentences with a postverbal subject would be left with the same lexical items as, in this framework, the numeration underlying sentences with a preverbal subject. And, what is worse, in fact one could never obtain a preverbal subject since, as sentences with postverbal subjects show, there is nothing compelling the subject to raise. Nevertheless, such a defeating conclusion is not warranted. Let us see why. In the first place, the disappearance of proEXPL does not imply as a inescapable consequence that the numeration of a sentence with a preverbal subject must be identical to that feeding a sentence with a postverbal subject. To obtain two different numerations without resorting to proEXPL, it suffices to reinterpret, in minimalist terms, the empirically well grounded assumption that in NSLs the so called preverbal subject is in fact left-dislocated, thus associated to a sentence-internal pro, 104 CatWPL 8, 2000 Joana Rosselló 10. Certainly, erasure of Case as a [–interpretable] feature both in nominal receivers (DP/NP) and in assigners (T for Nominative; V for Accusative —and other inherent Cases) does not fit this pattern. We could argue, then, that the essential point in our proposal is that there must be an asymmetry between the two members of the checking match, one option being that one such an asymmetry is [–interpretable]/[+interpretable]. Another option, relevant for Case, would be the Case-asigner/Case-receiver asymmetry. This way of preserving the asymmetry becomes impossible in Chomsky (1998), where structural Case in nominals is taken to be the reflection of a set of [–interpretable] φ-features. CatWPL 8 001-137 29/5/2001 17:30 Página 104 whereas the postverbal subject is in its θposition. This reinterpretation could very easily adopt the following form: the preverbal subject is indeed a non argument DP merged above the θdomain, either in the Inflection phase (like a true expletive) or above, in the Complementizer phase (like a true left-dislocated element) —see footnote 19. If it is like an expletive, it has to have an associate that is the true argument. In this case, however, the associate is not lexical, but a pro occupying a θposition in vP. Notice that under this view a sentence with a preverbal subject must have a pro in the numeration, whereas a sentence with a postverbal subject cannot have any. Both pro (either associated with a preverbal subject or not —in a null subject sentence) and lexical postverbal subject appear in their θposition. To conclude this first rough approximation to the present proposal, it is worth noticing that it has a clear advantage over previous ones assuming the leftdislocated nature of preverbal subject in NSLs. In effect, unlike its predecessors, the present proposal need not resort to additional structure to state the non argumental nature of the preverbal subject. Under the derivational view of MP, this nature can be stated through the derivation itself. In accordance with what we have just seen, it appears that it would have been the wrong move to conclude that the interplay between preverbal and postverbal subject in NSLs can not follow from the very workings of CSHL. Moreover, it is in fact a surprising outcome that even in this framework the idea that the preverbal subject is somehow extraneous to CSHL can be expressed. Upon careful consideration, we will see that although the preverbal subject enters the derivation as an expletive (when merged with T, at least) it does not fully work as such: an expletive has to check the D-feature of T, while our preverbal subject does not. The checking/erasure of the φ-features of V and of the categorial D/N-feature of T is always achieved covertly by the features of the DP/NP postverbal subject. In this connection, we have to bear in mind that when in the course of the derivation we derive T, nothing more is needed in NSLs: irrespective of the additional merging of our preverbal (expletive) subject, the derivation would converge. In light of this, it seems sensible to conclude that at this point, T (T0) (in NSLs), has all the features required for the derivation to converge. And the ultimate reason for this conclusion is of course that in NSLs there is an argumental pro, as suggested by plenty of evidence. It is a nice result, as I have just said, that the optional nature of the preverbal subject can be accounted for by the very mechanisms of CSHL. It could be argued, however, that the argument just developed is fallacious because in NSLs not only the preverbal subject is optional but also the postverbal one. So it seems at first glance, at least for NSLsUA. For NSLsAA, instead, the argument stands, since in these languages the mere dropping of a postverbal subject would leave V deprived of φ-features (according to U33 of Greenberg) and thus unable to legitimate a null subject (pro). Should we conclude that the argument in the previous paragraph can apply only to NSLsAA? Not at all: the argument, contrary to appearances, is also valid for NSLsUA. It is a fact, which has gone unnoticed, that in NSLsUA a postverbal subject can not always be left out, whereas a preverbal one always can. More specifically, there are sentences that become ungrammatical when deprived of theirs postverbal subject, whereas nothing parA Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 105 CatWPL 8 001-137 29/5/2001 17:30 Página 105 5. A minimalist approach to the Null Subject Parameter The purpose of this section is to characterize, as succinctly as possible, the essential plot of the hypothesis defended here, so as to compare it with the essential plot of the (by now classical) proposal, according to which NSLs are just like English, except for the possibility of postverbal subjects and the availability of a phonetically null pronoun.20 In order to establish the comparison more clearly, I will make use of a minimally formalised presentation. 5.1. Two conceptions. A comparison The idea I am putting forward, as implemented in the framework of Chomsky (1995), would be the following: (9) a. Preverbal subj. [TP DP [T 0max FFS– V+φ–T] [vP proStv ...] ] Ex.: En Joan ballava. the Joan danced b. Postverbal subj. [TP [T 0max FFS– V+φ–T] [vP DPStv …] ] Ex.: Ballava en Joan. danced the Joan 112 CatWPL 8, 2000 Joana Rosselló (i) a. *Que en Pere ho sap? WH the Peter it-knows ‘Does Peter know?’ b. Que ningú ho sap? WH nobody it-knows ‘Doesn’t anybody know?’ (ii) a. *Ja en Pere ho sap. already the Peter it-knows ‘Peter already knows’ b. Ja ningú ho sap. already nobody it-kows ‘Nobody knows anymore.’ In view of this facts, I would propose that in Ningú queda (nobody stays/*is-left) or Ningú falta (Nobody is-absent/*is-left) preverbal ningú, like the above mentioned DPs, is directly adjoined to a root node of the sentence (hence the only agentive reading of the preverbal position), but, unlike DPs, it would not be adjoined to T, but to NEGation, which would be universally lower than T, according to the evidence from languages where negation is not a verbal clitic. Admittedly, this amounts to saying that strong V-features of T fail to be checked. I would say, however, that, in accordance with the discussion in section 3 (and footnote 7), checking only requires that the verb must be outside the vP (θ-domain), and move into the checking domain, so that merge with NEG would satisfy this requirement. If this reasoning is on the right track (which would need much more careful examination), we should further refine the idea that NSLs have a strong V-feature in T. Tentatively, we could assume that in NSLs the first inflectional category above vP has a strong V-feature. Notice that such a solution is in accordance with the philosophy of the MP, in the sense that requirements must be satisfied locally and in the most economical way. 20. Here I will ignore V-movement. CatWPL 8 001-137 29/5/2001 17:30 Página 112 c. Null Subj. [TP [T 0maxFFS– V+φ–T] [vP proStv …] ] Ex.: Ballava. danced [FF: Formal Features] We appeal to Merge to derive the preverbal subject, and to Procrastinate, hence to Move F(eatures), to derive both the postverbal subject and the (also postverbal) null subject. Let us now compare (9) with the ‘English-like’ version (10): (10) a. [TP subjecti[T’ V [VP titV]]] b. [TP proEXPL [T’ V [VP DP tV]]] c. [TP proi[VPti .. ]] [traces, indices and bars are a notational convention herein as in (9)] Now, even if we were to make the assumption (which nevertheless can be proved to be clearly false) that the empirical side does not favour either option, the favoured option from the point of view of the computational operations required to derive the triad would no doubt be the one defended here, namely (9). Let us see why. In (10), both the preverbal subject and the null subject (here taken to be preverbal) require the application of Move in overt syntax, an operation that is complex and hence «last resort» —in Chomsky (1998)21 it is the result of combining the two other possible operations Merge and Agree—, while it can be dispensed with in (9). The only movement in (9) is merely one of features (Move F), satisfying Procrastinate —in Chomsky (1998) Move F becomes Agree, without movement of any kind. 5.2. Merge The proposal in (9) introduces an occurrence of Merge involving a DP which, despite not being an expletive, is purely merged with T. This application of Merge has the following properties: a) it does not violate any requirement, since the DP is non-argumental; it therefore complies with Chomsky’s (1995, 1998: (6)) point that «pure merge [merge which is not part of move] in θ position is required of (and restricted to) arguments»; b) it is preferable because it renders unnecessary the more complex option Move; c) it establishes a relation with its correlate pro which is not part of the checking mechanisms proper; actually, neither the one nor the other contains any [–interpretable] features, which in Chomsky’s (1998) terms is a necessary condition for a checking relation to be established —we are now assuming that eliminating Nominative Case in NSLsUA is the right move. This would reasonably provide an account for the fact that this DP is always an optional addendum as far as strictly computational operations are concerned; d) it captures the well known differences between preverbal subjects in NSLs and English-style lanA Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 113 21. I have chosen to use the theoretical machinery in Chomsky (1995: Chap. 4), since the proposal in Chomsky (1998) is not well-defined as it stands. See Rosselló (1999). CatWPL 8 001-137 29/5/2001 17:30 Página 113 guages, by attributing them to the derivation and not to the representation. Against a certain classical representational view that held that the preverbal subject is dislocated, outside C’/CP, the derivational view proposed here allows us to dispense with superfluous structure, since the preverbal subject in (1a) would occupy the same position as the English subject;22 e) it opens up new lines for the understanding of properties and phenomena that have proved resistant to treatments that assume that the preverbal DP in (1a) is obtained by movement. So, for instance, (i) inversion effects in interrogatives —a still unsolved topic for NSLsUA— could be dealt with as a case of relativized minimality; (ii) more satisfactory paths could open up to deal with cases of «approximate» agreement (Els lingüistes no sabem res —the linguists not know-1PL nothing ‘We linguists do not know anything—; Els estudiants ho fem tot nosaltres —the students it-do-1PL everyting we ‘We estudents do everyting ourselves’—; (Nosaltres)ho fem tot les dones —(we) it-do-1PL everything the women ‘(Among us) we women do everything’—) and to shed light on the control and binding behaviour illustrated in examples like Els joves ho aconseguireu sense esforçar-vos/*se gaire —the young it-achieve-FUT2PL without to-strain-yourselves/*themselves much ‘You the young will achieve it without much effort—; (iii) a well founded explanation could be provided for Greenberg’s universal 33 (hence for the characteristic pattern in NSLsUA): the preverbal subject cooccurs with maximally specified verbal agreement, precisely the agreement that shows up with null-subject sentences in those languages; this would follow if this «subject», unlike postverbal ones, is a DP merged with a sentence containing a phonetically empty pronoun; and finally (iv) it provides a better understanding for the «loss» of interpretations (or even of grammaticality, in some cases) triggered by preverbal (as opposed to postverbal) subject position in certain constructions (involving a copulative verb—see (2c) vs. (3c); o an unaccusative verb —see (2a) vs. (3a)). In what follows, I will only minimally develop part of this last point, which was already introduced in sections 3 and 4. 6. Impossible preverbal (or null) subjects In § 3 and 4.2 we had occasion to see that preverbal non focused subjects (or null subjects) enjoy less freedom of appearance than postverbal (or focused/wh preverbal) ones (see (2), (3), (7) and (8)). To start with, such a situation is incompatible with the idea that the preverbal subject is an Sv-internal subject that has moved (an almost inescapable analysis in a framework assuming the Sv-Internal Subject Hypothesis)23. The data, however, have taken us much further. Specifically, the 114 CatWPL 8, 2000 Joana Rosselló 22. Merge can take place in higher or lower positions, though. See note 19. 23. There are approaches that do not challenge this point, which simply follows from adopting the Internal Subject Hypothesis, and still hold the view that the postverbal subject as an inverted subject. How? By deriving it from a preverbal subject. In this kind of approach (Kato 1999) there is a preverbal subject originating as an Sv-internal subject; and there is an overt postverbal subject which is the result of inverting the derived preverbal subject. Besides the self-evident complication, the problem is that this framework also predicts, contrary to fact, that between a preverbal and a postverbal subject there cannot be any more differences than those relative to information packaging. CatWPL 8 001-137 29/5/2001 17:30 Página 114 identical behaviour, in all such cases, of postverbal subjects and focused/wh preverbal leads us to an obvious conclusion: if focusing of the preverbal subject is the only strategy for preserving the postverbal reading of the subject, then Nominative Case cannot be available for the variable bound by the focus/wh operator if, in accordance with the orthodox view, the Nominative Case position is the preverbal subject position. To take an example, we could not expect that a couple of Majorcan sentences like Qui queda? (‘Who stays?’ / ‘Who is left?’) and Na MARIA queda (‘MARIA stays’ / ‘Maria is left.’) could have the stative-quantificational reading (‘Who is left?’, ‘MARIA is left!’), besides the agentive one (‘Who stays?’ ‘MARIA stays!’) —which in Barcelonan involves a pronominal verb: Qui es queda?, La MARIA es queda!—, while the stative-quantificational reading does not obtain in Na Maria queda (‘Maria stays’ / *‘Maria is left.’), which can be interpreted only agentively —and is simply impossible in Barcelonan: *La Maria queda. Notice that there is no way to escape this conclusion. It would not be legitimate to assume that wh movement of the subject takes place from the θ-merge basic position24 and that then covert checking of Nominative Case applies to the trace (copy). To be precise, this escape-strategy would not be such since it would only be applicable to the instances where it would not be necessary (checking could also apply overtly, economy considerations aside): it would apply only in cases where the purported preverbal Nominative Case position is in fact associated to the wh reading. The reason for this limitation is this: covert raising to merge DP features with the T inflective head places these checking features in a structural position which must have essentially the same properties as the [Spec, T] position. And this leads again to the wrong prediction that if we get the non-agentive reading of Qui queda? we should also get it for Na Maria queda. Now, this conclusion follows from the view that assimilates covert movement of DP/NP features to overt movement of DP/NP, which is precisely the position taken by Chomsky (1995: § 4.4.4), in consistence with his background assumption that the Spec,T position is formed by either a subject DP raised from the vPinternal θ-position, or by merge of an expletive. If we keep to this assumption, the assimilation is conceptually plausible and empirically tenable. What is not tenable, however, is the assumption itself, namely that the only way to form a Spec,T is the one contemplated in Chomsky (1995): (11) Spec of T is created by raised DP or an expletive merged with T. If (11) was not the only option, further possibilities could be explored to get Nominative Case in sentences of the kind Qui queda? and La MARIA queda in the stative-quantificational interpretation, so as to preserve the idea that an argument is a non-trivial Chain (Case, θ). Let us see why (11) cannot be the only option. A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 115 24. It is most remarkable and symptomatic that from a variety of data and of independent theoretical constructs (the Empty Category Principle in the Government and Binding period and the Checking theory in the MP derivational approach) one is forced to the conclusion that wh extraction in NSLs takes place from a postverbal position. CatWPL 8 001-137 29/5/2001 17:30 Página 115 If (11) was the only option, it would be legitimate to assimilate, in the way suggested above, the position of adjunction to the T head with the Spec,T position, and therefore a postverbal subject (always associated to an expletive in Spec,T) should have the same properties as a preverbal subject, since the features of the expletive-associate (the postverbal subject) would merge with T covertly. In my view, this assumption is directly challenged by the pairs (2)-(3) and (7)-(8). And, if we go further to test the possibilities for the subject to act as a controller —the test used by Chomsky (1995)—, it does not seem to me tenable that in NSLs the preverbal or postverbal subject position does not make a difference: control is possible in (12) while it sounds unnatural in (13): (12) a. En Pere ve per controlar. the Peter comes to control b. Un pacient ha entrat per mirar. a patient has come-in to look (13) a. ?*En vénen dos per fer el sopar(, de cuiners).25 EN-come two to make the dinner (of cooks) b. ?* No n’entra gaire a mirar l’exposició (, de gent). not EN-come-in many to look-at the-exhibition (of people) c. ?* N’han arribat un munt per estudiar el tema (, d’experts). EN-have arrived a lot to study the topic (of-experts) Contrary to the uniform evaluation that has been given in the MP to Cardinaletti’s examples adduced by Chomsky (1995: 274), my view is that it makes a difference whether the postverbal subject involves extraction of clitic en (it. ne) or it stays in full. In the latter case, in fact, the examples sound less unnatural: (14) a. Vénen dos cuiners professionals per fer el sopar. come two cooks professional to make the dinner b. No entra gaire gent a mirar l’exposició. not come-in many people to look-at the-exhibition c. Ha(n) arribat un munt d’experts per estudiar el tema. have arrived a lot of experts to study the topic In this connection, the following pair —from Majorcan— is also very illuminating: 116 CatWPL 8, 2000 Joana Rosselló 25. One can devise even more deviant examples. In a sentence like No n’hi ha prou/gaires per fer el sopar (not EN-THERE-have enough/many to make the dinner ‘There are not enough/many of them to make the dinner’), the clitic en cannot be the controller of the subject in the purpose clause, which can only have an arbitrary reading. En, in such sentences, points to another referent (guests, kind of food, etc.) which can be interpreted as a prerequisite for the action expressed in the purpose sentence. CatWPL 8 001-137 29/5/2001 17:30 Página 116 (15) a. No en queda cap per visitar (de pacient)/ (?*de metge). not EN-is-left/*stays any to visit (of patient)/ (?*of doctor) ‘(As for patients) none is left to be visited.’ *‘(As for doctors), none stays to do their visits.’ b. No queda cap pacient/ cap metge per visitar. not is-left/stays any patient / any doctor to visit ‘No patient is left to be visited’ ‘No doctor stays to do their visits.’ In (15a), where extraction of clític en renders the unaccusative structure the only option for the main clause, the only available interpretation is the stative-quantificational one, and subject control of the infinitive becomes impossible: the main clause argument is interpreted as coreferent with the object of the infinitive, so that en is interpreted as «patient» and not «doctor», as shown by its dislocated correlate —on the pragmatically unmarked reading. In (15b), instead, where clitic en is not extracted, the main clause structure is ambiguous: on the stative-quantificational reading, it behaves as in (15a), with cap pacient as the pragmatically appropriate option for the V sister position; on the agentive (‘stay’) reading of quedar, instead, subject control of the infinitive becomes available, the infinitival verb being used intransitively. In this case, of course, the pragmatically preferred interpretation chooses cap metge as subject for unergative quedar (‘stay’). This behaviour is not idiosyncratic of Majorcan, as evidenced by the ambiguity of standard Catalan sentences like Abans faltaves tu i ara falto jo (‘Before, YOU were missing/absent and now I am’), Ja no falta ningú (‘Nobody is missing/absent anymore’), Ara falta la Maria (Now MARIA is missing/absent’). Let us now consider the contrast in (16): (16) a. (Dos individus) *han vingut per veure. (Two guys) have come to see b. Han vingut dos individus per veure. Have come two guys to see ‘Two guys have come to be seen.’ From (16a) it can be inferred that a preverbal or null subject is unable to provide the content of the obligatory object of veure ‘see’ (in Catalan, the stative intransitive interpretation of veure —namely ‘be able to see’—requires the clitic hi: veurehi) and thus that control is obligatory: the preverbal (or null) subject cannot fail to be a controller. This is not the case, instead, with (16b), where dos individus can be the antecedent of the infinitive object.26 We have, in sum, a further example of differential behaviour depending on the presence of a preverbal or a postverbal subject, namely a further counterexample to assimilating one to the other under (11). A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 117 26. The status of (16b) is claimed to be dubious by a reviewer of this paper. The contrast between (16a) and (16b) holds, in any case. CatWPL 8 001-137 29/5/2001 17:30 Página 117 The position defended here is well known: (11) is not the only possibility for creating Spec,T. Specifically, NSLs can create it by merging with T a DP correlated with the pro element, which stays in Spec,vP position. From this standpoint, new paths open up to us to explore new possibilities on how Nominative Case can be obtained —remember that this possibility is linked to the non-universality of (11). It also remains to be argued for that the position for pro is Spec of vP. In what follows, we will try to tie up these two issues (Nominative Case and pro position) and, correspondingly, make sense to the above data, going beyond the mere assumption that the preverbal subject position is obtained by direct merge with T. Consider the sentences in (17) (where the verb quedar is intended in the Majorcan use: ‘be left’/‘stay’). (17) a. Quedes tu. / Faltes tu. stay/are-left you are-absent/are-missing you a’. Queda na Maria. / Falta la Maria. stays/is-left the Maria is-absent/is-missing the Maria b. Tu quedes. / Tu faltes. you stay/*are-left you are-absent/*are-missing b’. Na Maria queda. / La Maria falta. the Maria stays/*is-left the Maria is-absent/*is-missing c. Quedes. / Faltes. stay-2SG/*are-left-2SG are-absent-2SG/*are-missing-2SG The sentences in (17a/a’) are ambiguous —they allow both the stative-quantificational (unaccusative) reading and the agentive (unergative) reading—, while those in (17b/b’), with a non-focused preverbal subject only allow the agentive (unergative) reading. The latter situation is also true of (17c). What is to be emphasised is that in (17a/a’) the unaccusative interpretation is not linked to the occurrence of clitic en extraction on the postverbal subject, nor even to the quantificational or bare-NP status of the postverbal subject: the unaccusative interpretation is possible with any DP, even a pronominal one, provided it is phonetically overt. In a way, the result is again surprising: why should a null subject (pro) not be able to licence the unaccusative interpretation, while a overt pronominal subject is? Should we perhaps appeal to an extragrammatical factor, to the effect that the unaccusative interpretation should be necessarily associated to focus interpretation on the argument? This is unlikely: a sentence like En queden (EN-are-left ‘Some are left’) is perfectly well-formed, and it only has the unaccusative interpretation, without any focusing on the argument.27 Let us see if we can do any progress on the side of Case theory, which is the other unsolved problem we are left with. If the unaccusative interpretation stems from 118 CatWPL 8, 2000 Joana Rosselló 27. It does not seem of any help to attribute to lexical pronouns a special and exclusive status —as in Rigau (1987), who treats them as quantifiers in A’ position— in order explain the facts in (17). The reason is that here lexical pronouns behave like nominal DPs —(17a) behaves like (17a’) and (17b) behaves like (17b’). CatWPL 8 001-137 29/5/2001 17:30 Página 118 the V-sister position, while the unergative one stems from the Spec,vP position, we could possibly appeal to the incompatibility of the former position with Nominative Case, which is required by pro in (17c). Now, this does not seem a sensible option when the sentences in (17a) themselves are unaccusative —in one of their interpretations— and their subjects feature in Nominative Case, at least apparently. If resorting to Nominative Case is of no use to explain the contrast between (17a/a’) —with two readings— and (17c) —with only the stative-quantificational reading— and, besides, the only plausible means of checking Nominative Case throughout (17) is to be covert raising of the DP features to T (which, in fact, are independently required by the φ-features of V in T), it is tempting to explore what would happen if in NSLsUA we dispense with Nominative Case throughout. Since it would be a more economical option, to the extent that we dispense with one (apparently superfluous) mechanism, that of Nominative Case checking, it seems worth of consideration. 6.1. On the spurious character of Nominative Case in NSLsUA Once considered seriously, elimination of Nominative Case in NSLsUA seems desirable. It would appear that Nominative Case is superfluous in a language where the [–interpretable] φ-features of the finite verb raised to T will always attract the features of the postverbal subject, whether an overt DP or pro. In other words, in a NSL Nominative Case is not necessary, while in languages like English Nominative Case is a necessary resource in addition to the EPP feature (or strong D-feature of T). It is, in effect, necessary since in constructions with a postverbal subject, which is the associate of a pure expletive (such as there), the only means of systematically achieving a convergent derivation is assuming that the expletive consists solely of the categorial feature D and that the associate, instead, comes equipped with φ-features and Nominative Case —both necessary.28 If, instead, there are no expletives, and V raised to T has [–interpretable] φ-features, Nominative Case would become redundant, since raising of the formal features of the postverbal subject and, consequently, subject agreement are guaranteed. At a more conceptual level, eliminating Nominative Case in NSLsUA is a desirable move too. Nominative Case is the expression of a Spec-head agreement relation. Now, if the relation between X and a DP (with FF1) in Spec of X is equivalent to the relation [XFF1-X] (in the case of pure feature checking without pied-piping), «real» (morphologically realized) agreement in φ-features (as in Venim nosaltres/pro —come-1PL we/pro—, Vénen els nens/pro —come3PL the children/pro—, Ho dius tu/pro —it-say2SG you/pro) can be considered an implementation of this abstract realization in its most economical format, namely pure feature checking by adjunction to the T head without pied-piping ([T FF1-T]). One further virtue of this view is that it does not do any violence (it rather lends support) to the idea that an argument always forms a non-trivial chain with its head A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 119 28. See the corresponding argumentation in Chomsky (1995: § 4.5.3) CatWPL 8 001-137 29/5/2001 17:30 Página 119 in a checking position with [–interpretable] feature(s) —in the case under consideration the φ-features of V— and its foot in a θ-position . Finally, it is to be pointed out that mere φ-feature raising gives the same effect as the EPP in a covert way: raising of D/N29 to T.30 More speculatively, it could be the case that, rather than the EPP —in Chomsky’s (1995) formulation—, what was obligatory was the transmission of predication to the inflectional stretch. Now, such a transmission can be achieved both in the configuration [TP DP T] (standard EPP) and in the configuration [TFFD/N Τ]. Without Nominative Case, sentences like Qui queda (who stays/is-left) and Na MARIA queda (The MARIA stays/is-left) can be derived by direct movement of the qu/focus operator. This movement can be originated either in the Spec,vP position (unergative reading) or in the sister-of-V position (unaccusative reading). The operator-variable (operator-argument) configuration would be derived in two steps: first, by overt raising we get the operator; second, by covert raising of the φ-features of the copy to the [–interpretable] φ-features of the verb, we get the argument. Interestingly, there is some independent evidence in support of dispensing with Nominative Case in NSLs. We will consider it after completing the analysis of the sentences in (17) and dealing in particular with the issue of the interpretation of (17c). 6.2. On the position of the null subject (pro) Following the conclusion reached in the preceding section, in the examples in (17) there is no Nominative Case. Let us consider how the corresponding derivations would proceed at the relevant stage. In (17a/a’) the double interpretation is obtained depending on which the θ-position is for the argument (la/na Maria / tu): if it is Spec,vP, the interpretation is agentive —in Barcelonan, as pointed out, this requires the pronominal form of the verb, quedar-se; if it is the V-complement position, the interpretation is stative-quantificational. As for the feature-checking mechanism, there should be no problem: φ-features of V covertly attract the φ-features of the argument, whether in one or the other position. The cases in (17b/b’) and (17c) should in principle be analogous to those in (17a/a’), the only difference being that now it is pro —a Caseless amalgam of the categorial feature Dand φ-features (person, number, gender)— that must check the φ-features of V in T.31 We are faced, however, with the already noted contrast: 120 CatWPL 8, 2000 Joana Rosselló 29. If the orthodox EPP amounts to the strong D-feature of T, the version defended here would not be completely equivalent: on the basis of sentences like Passen coses (happen things), Falta cafè (islacking coffee) (see also the analysis of (20a), (20c)), it can be perfectly argued that it is an N (NP) and not a D (DP) that erases the number features of the raised verb (in these cases, third person can be considered a non-person). 30. Or to an inflectional (checking) category higher to that of θ-merge. See footnote 19. 31. It is to be pointed out that these data do no seem to be explainable from approaches that consider agreement as an argument (or at least a determiner), such as Ordóñez & Treviño (1999), Kato (1999), Alexiadou & Anagnostopoulou (1998) etc. They all seem to predict a uniform behaviour throughout (17), since agreement does not vary at all across the different cases in (17). CatWPL 8 001-137 29/5/2001 17:30 Página 120 both (17b/b’) and (17c) can only have the agentive interpretation, that we associate with the Spec, vP position. Does this mean that pro can only yield convergent derivations in this position? That’s what we should infer. Let us consider what would happen otherwise. A sentence like Els nens miren (the children watch, ‘The children are watching’), which contains a verb that does not obligatorily carry Accusative Case (and so can yield an absolute, intransitive interpretation), should be ambiguous between this absolute interpretation and the one yielded by the sentence Els nens són mirats (the children are watched),32 contrary to fact. The problem that arises if the distribution of pro is not restricted is that we would have no means to know whether it is present or absent in a given derivation. In a word, a pro, exclusively consisting of [+interpretable] features from the point of view of the C-I system, but phonetically empty, must be identifiable in an optimal way. And this implies that it must be identifiable unequivocally from the point of view of the interpretation and locally from the point of view of the feature-checking relations that it is involved in and identified by. Now, the local identification requirement would amount to occupying the highest position in the θ-domain, i.e. Spec,vP. It is impossible to review here all the consequences of the present proposal.33 Let me just say something on its possible consequences on how we view the relation between an object clitic (direct object and dative) and its corresponding θ-position. I would like to point out that, according to what the SLH establishes, clitics should be kept a part from inflection because of the simple A Minimalist Approach to the Null Subject Parameter CatWPL 8, 2000 121 32. More explicitly, if mirar (‘look’, ‘watch’) has (or can assign) Nominative Case and pro is Caseless, starting from a numeration with the items miren¸ pro, T and C we could both get an unergative vP, [SVpro miren], and an unaccusative VP, [SV miren pro], both ultimately converging. It turns out, though, that only the former derivation converges, since miren only yields the unergative interpretation. The other one, [SV miren pro], which would be interpretatively equivalent to a passive —are watched— without passive morphology, is not obtained. How can we prevent [SV miren pro] from converging? What we propose here is that language design must limit pro to Spec,vP, precisely to avoid the indeterminacy that otherwise would arise. 33. This proposal on the distribution of pro implies that unaccusative sentences with a null (non partitive) subject have pro in Spec,vP. I don’t think this is necessarily a problem. I would consider exploring the following idea: lexically unaccusative verbs can unergativize in the syntax (by means of raising by substitution to Spec,vP) This option would become obligatory when the argument is pro. This would account for the very clear contrast in agentivity perceptible between Vénen (come3PL , ‘They come’) and En vénen (EN-come-3PL, ‘Some come’) —cf. Vénen per veure’t (‘They come to see you’) vs. *En vénen per veure’t (‘Some come to see you’). Overt DPs (excluding bare NPs), instead, apparently would be able to appear both in Spec,vP and in V-sister position. (see (17a/a’). In fact, one could speculate that, contrary to what has been previously assumed, the Majorcan verbs faltar (‘be missing’/‘be absent’) and quedar (‘be left’/‘stay’) are not lexically ambiguous: rather what happens is that the process of syntactic unergativization yields as a result a more neat contrast on the interpretative side. The reason could lie in their lexical meaning. So, the meaning of unaccusative faltar implies «non presence», while that of unergative faltar would imply «voluntary, controlled non presence». The fact that in non Majorcan Catalan (as in Spanish) quedar does not allow unergativization would have the following interpretation: unergativization involves a movement operation that can be spared if another lexical item is available, in this case the pronominal verb quedar-se, which is inherently unergative. CatWPL 8 001-137 29/5/2001 17:30 Página 121 Levin, B.; Rappaport Hovav, M. (1995). Unaccusativity. Atthe Syntax-Lexical Semantics Interface. Cambridge, Mass.: MIT Press. Longobardi, G. (1987). «Las oraciones copulativas en la teoria sintáctica actual». In Demonte, V.; Fernández, M. (eds.) (1987). Sintaxis de las lenguas románicas. Madrid: El Arquero. Manzini, R.; Savoia, L.M. (1997). «Null subjects without pro». WPLUCL 9: 303-313. Moro, A. (1997). The Raising of Predicates. Predicative Noun Phrases and the Theory of Clause Structure. Cambridge: Cambridge University Press. Nash, L.; Rouveret, A. (1996). «Functional heads and Proxies: an asymmetric set». Glow Newsletter 36: 50-51. Ordóñez, F.; Treviño, E. (1999). «Left dislocated subjects and the pro-drop parameter: A case study of Spanish». Lingua 107: 39-68. Picallo, C. (1998). «On the Extended Projection Principle and the Null Expletive Subject». Probus 10: 219-241. Piera, C. (1987). «Sobre la estructura de las cláusulas de infinitivo». In Demonte, V.; Fernández, M. (eds.). (1987). Sintaxis de las lenguas románicas. Madrid: El Arquero. Platzack, C. (1994). «Null Subjects, Weak AGR and Syntactic Differences in Scandinavian». WP in Scandinavian Syntax 53: 85-106. Pollock, J.-Y. (1997). Langage et Cognition. Introduction au programme minimaliste de la grammaire générative. Paris: Presses Universitaires de France. Rigau, G. (1987). «Sobre el carácter cuantificador de los pronombres tónicos en catalán». In Demonte, V.; Fernández, M. (eds.). (1987). Sintaxis de las lenguas románicas. Madrid: El Arquero. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. — (1990). «Residual Verb Second and the Wh-criterion». Technical Reports in Formal and Computational Linguistics. University of Geneva. Also in Belletti, A.; Rizzi, L. (eds.) (1996). Parameters and Functional Heads: Essays in Comparative Syntax. Oxford: Oxford University Press. Rosselló, J. (1999). «Un enfocament minimista del Paràmetre del Subjecte nul». Reports de Recerca del Grup de Gramàtica Teòrica de la UAB, GGT-06 (http://ggt.uab.es). Solà, J. (1992). Agreement and Subjects. Universitat Autònoma de Barcelona. Doctoral thesis. Taraldsen, K.T. (1980). On the Nominative Island Condition, Vacuous Application, and the that-trace filter. Indiana University Linguistics Club. 128 CatWPL 8, 2000 Joana Roseselló CatWPL 8 001-137 29/5/2001 17:30 Página 128